A former national security official who served on the National Security Council and at the Pentagon during the first Trump administration is warning that artificial intelligence could become the next strategic chokepoint where the United States depends on China — and that the danger may come from an unexpected direction: America’s own regulatory instincts.
In an essay published by The Daily Wire, Mike Martin argues that Washington’s AI debate is focused on the wrong question. Policymakers and pundits keep asking which country has the best model and how close Chinese labs are to the frontier, he writes. But that framing treats AI as a finished product — a skyscraper — when in reality it is becoming something more like steel or a rare-earth magnet: a component that disappears inside everyone else’s products.
That distinction, he contends, matters enormously for how the U.S. handles open-weight models that companies can download, customize, fine-tune, run locally, and embed without paying a frontier lab for access.
The rare earths playbook, repeated
Martin walks readers through a familiar history. For decades, American firms followed a simple rule: buy from whoever can make it cheaper. That logic played out in rare earths, pharmaceutical supply chains, and steel. Consumers benefited, costs fell, and capital flowed toward more profitable industries. Only later did it become clear that cheap access to a strategic resource is not the same as secure access.
The U.S. was once the world’s dominant rare earths producer, with California’s Mountain Pass at the center of the industry, Martin notes. Beginning in the 1990s, inexpensive Chinese production changed the economics. Mountain Pass’s separation operations closed in 1998; mining stopped in 2002. Today, he writes, the U.S. gets more than 90 percent of its separated rare earths from China or from countries using Chinese material.
No American leader ever formally decided to depend on China for materials essential to electronics, weapons, and advanced manufacturing, according to Martin. The outcome emerged from thousands of individually rational purchasing decisions.
Steel followed a similar path, he says, with China expanding production under rules that permitted mass pollution, cheap labor, and currency manipulation, depressing global prices and pushing domestic producers out of business. Pharmaceuticals have gone the same way, with active ingredients and their chemical precursors increasingly sourced from China and India.
Those vulnerabilities were thrown into sharp relief during the COVID-19 crisis, when Americans confronted shortages of essential goods, delays for commonplace products, and sharply rising costs as supply chains buckled. They are back in the spotlight now, Martin writes, as the U.S. tries to secure the minerals needed for military systems, automobiles, and computers — a subject that he expects to arise when Chinese President Xi Jinping visits the White House.
Why open-weight AI is the test case
Applied to AI, Martin argues, the same dynamic is already taking shape. Intelligence is turning into a component. You don’t see the model running behind your finance software, your navigation system, your email inbox, or your phone’s camera — those tools simply work better because of it. For the companies building them, the question may eventually be less about who has the smartest model and more about who sells the cheapest intelligence that reliably does the job.
That’s why open-weight AI is the crucial battleground, in his telling. Businesses want models they can run on their own hardware, adapt to their needs, and embed in products without paying a frontier lab for every use. If American companies can’t supply those models, someone else will.
China, he writes, is increasingly willing to compete aggressively in exactly that market. Chinese developers have released capable open-weight models while stressing efficiency and low prices. They don’t need the world’s best model to reshape the market. They only need one that is good enough and cheaper.
Martin then poses a scenario: imagine the U.S. restricts open and affordable AI models, confining its economy to regulated, closed frontier systems. That, he argues, is precisely how industrial dependencies form. An American company could own the interface, the customer relationship, and the application — while the foundational intelligence underneath comes from a Chinese model, much as an American drugmaker can sell a pill whose critical chemical inputs originate overseas.
Notably, he writes that a straightforward ban on Chinese AI would not by itself solve the problem. If American open-weight companies aren’t encouraged to develop and compete, China will keep producing and dominating globally regardless.
America, he concludes, should think carefully before regulating itself out of offering the kind of intelligence the rest of its economy actually wants to build with — the lesson of rare earths, steel, and pharmaceuticals being that letting China flood a market with a cheaper product tends to end with China dominating that market. The most consequential input of the 21st century, he warns, is not the place to repeat the mistake.
Martin’s byline identifies him as a former senior national security official who served in the National Security Council and the Pentagon during the first Trump administration. His essay appears as part of a broader, ongoing debate over how far Washington should go in restricting access to foreign AI models — and whether the most effective answer lies in restrictions, in competition, or in some combination of the two.
Source: www.dailywire.com — https://www.dailywire.com/news/americas-next-china-dependency-could-be-artificial-intelligence